🔢 mathematics

Conditional Expectation expression in mean-field SDEs and its applications

This paper introduces a novel formulation of conditional expectations for jump-diffusion mean-field stochastic differential equations using Malliavin calculus and directional derivatives, which significantly improves the numerical pricing accuracy of American put options compared to conventional methods.

Original authors: Samaneh Sojudi, Mahdieh Tahmasebi

Published 2026-02-17

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Samaneh Sojudi, Mahdieh Tahmasebi

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to predict the future price of a stock, but this isn't just any stock. It's part of a massive, interconnected economy where every investor's decision affects everyone else. It's like a crowded dance floor: if one person steps left, the whole crowd shifts. In math and finance, this is called a "Mean-Field" system.

Now, imagine that on top of this crowded dance floor, there are sudden, unpredictable earthquakes (jumps) that knock people over. This is a "Jump-Diffusion" system.

The paper you provided is about a new, super-smart way to calculate the value of "American Options" in this chaotic environment.

What is an American Option?

Think of an American Option as a coupon for a discount on a product (like a stock).

European Option: You can only use the coupon on a specific date (like Black Friday).
American Option: You can use the coupon anytime before it expires.

The tricky part? You have to decide when to use it. If you use it too early, you might miss out on a bigger discount later. If you wait too long, the discount might disappear. To price this correctly, you need to calculate the "Conditional Expectation": "If the stock is at price X today, what is the best thing to do tomorrow?"

The Problem: The "Curse of Dimensionality"

Usually, to figure out the best time to use the coupon, mathematicians try to map out every possible future path.

The Old Way: Imagine trying to draw a map of every possible path a drunk person could take through a city. If the city has 1 street, it's easy. If it has 10 streets, it's hard. If it has 100 streets (which happens when you have many assets or complex market interactions), the map becomes so huge it would fill the entire universe. This is the "Curse of Dimensionality." Traditional computers crash trying to solve this.

The Solution: The "Magic Weight" (Malliavin Calculus)

The authors of this paper developed a new mathematical "magic trick" using something called Malliavin Calculus.

Here is the analogy:
Imagine you are trying to guess the average height of people in a room, but you only want to know the average height of people who are taller than 6 feet.

The Old Way: You measure everyone, write down their height, and then throw away everyone under 6 feet. Then you average the rest. This is slow and wasteful.
The New Way (This Paper): Instead of throwing people away, you give everyone a special weight.
- If someone is 6'2", you give them a heavy weight (they count a lot).
- If someone is 5'8", you give them a tiny, almost invisible weight (they barely count).
- If someone is 5'0", you give them a negative weight (they cancel out the noise).

By using these "weights," you can take a simple average of everyone in the room, and the math automatically filters out the people you don't care about. You get the answer for the "tall people" without ever having to separate them physically.

How They Did It (The Ingredients)

The paper combines three complex ingredients into a delicious stew:

The Crowd (Mean-Field): They account for how the group's behavior changes the individual's path.
The Earthquakes (Jumps): They handle the sudden, random shocks (like a stock market crash) using a specific type of math called Poisson Space.
The Magic Weights (Malliavin Calculus): They invented a new way to calculate these "weights" specifically for this crowded, earthquake-prone environment.

Why Does This Matter?

The authors tested their method on American Put Options (a type of financial contract).

The Result: Their method was much more accurate and faster than the old methods.
The Analogy: If the old method was like trying to find a needle in a haystack by looking at every single piece of hay one by one, this new method is like using a magnet. It finds the needle instantly, even if the haystack is huge and shaking.

The "Variance Reduction" (Making it Even Better)

The paper also talks about "Variance Reduction."

Analogy: Imagine you are trying to hear a whisper in a noisy room.
- Standard Method: You shout "Can you hear me?" and listen. Sometimes you hear it, sometimes you don't, because of the noise. You have to shout 1,000 times to get a clear answer.
- This Paper's Method: They use a "noise-canceling headset" (called Localization). They tune the headset to filter out the specific background noise of the market. Now, you only need to shout 10 times to get a crystal-clear answer.

Summary

This paper is a breakthrough in financial mathematics. It gives us a new, highly efficient tool to price complex financial contracts in a world where:

Everyone influences everyone else (Mean-Field).
Sudden shocks happen (Jumps).
We need to make decisions at any moment (American Options).

Instead of getting lost in a maze of infinite possibilities, the authors built a smart shortcut that uses "mathematical weights" to navigate the chaos, saving time, money, and computer power.

1. Problem Statement

The paper addresses the computational challenges associated with pricing American options in high-dimensional financial markets where asset prices are influenced by mean-field interactions and jump processes.

Mean-Field Dynamics: In interconnected markets (e.g., energy grids, large portfolios), individual asset dynamics depend not only on their own state but also on the statistical distribution (law) of the entire system. This is modeled via McKean-Vlasov Stochastic Differential Equations (SDEs).
Jump-Diffusion: Financial markets often experience abrupt shocks (crashes, news events). The model incorporates Poisson jump processes alongside standard Brownian motion.
The Core Challenge: Pricing American options requires calculating conditional expectations at each time step to determine the optimal early exercise strategy (the "early-exercise feature"). In high dimensions ( $d > 3$ ), traditional deterministic methods (finite difference, finite element) fail due to the "curse of dimensionality." While Monte Carlo methods are feasible, estimating conditional expectations in jump-diffusion mean-field settings is mathematically complex and computationally expensive due to high variance.

2. Methodology

The authors propose a novel framework combining Malliavin Calculus on Poisson space with Mean-Field SDEs to derive exact representation formulas for conditional expectations.

A. Mathematical Framework

Model: The underlying asset $X_t$ follows a mean-field SDE with jumps:
$dX_t = b(t, X_t, \mathbb{E}[\phi(X_t)])dt + \sigma(t, X_t, \mathbb{E}[\psi(X_t)])dW_t + \int \lambda(t, X_t, z, \mathbb{E}[\xi(X_t)])\tilde{N}(dt, dz)$
where $\tilde{N}$ is a compensated Poisson random measure.
Malliavin Differentiability: The authors prove that the solution $X_t$ is Malliavin differentiable with respect to both the Wiener process ( $W$ ) and the Poisson process ( $N$ ). They establish the existence of the stochastic flow and derive explicit expressions for the directional derivatives ( $D_{r,z}X_t$ and $D_r^W X_t$ ).

B. Conditional Expectation Representation

The core theoretical contribution is converting the conditional expectation $E[f(F) | G = \alpha]$ into an unconditioned expectation weighted by specific Malliavin weights.

Integration by Parts: Using the duality formula of Malliavin calculus on Poisson space, they derive:
$E[f(F) | G = \alpha] = \frac{E[H(G-\alpha)(f(F)\Pi_1 + f'(F)\Pi_2)]}{E[H(G-\alpha)\delta(u)]}$
where $H$ is a smoothed Heaviside function, $\delta(u)$ is the Skorohod integral, and $\Pi_1, \Pi_2$ are specific weights derived from the system's coefficients.
Handling Non-Differentiability: For non-differentiable payoff functions (common in options), they utilize a localization technique involving a process $\hat{u}$ to handle the singularity, extending the representation to cases where $f$ is merely Borel-measurable.

C. Variance Reduction

To make the Monte Carlo simulation computationally viable, the authors implement variance reduction techniques:

Localization: They introduce a localizing function $\psi$ (optimal choice identified as a Laplace distribution) to minimize the integrated mean squared error.
Control Variates: The method utilizes the derived weights to reduce the variance of the estimator significantly compared to standard regression-based approaches (like Longstaff-Schwartz).

D. Algorithm for American Options

The authors propose a Malliavin Monte Carlo (MMC) algorithm:

Simulate $M$ paths of the mean-field SDE using the Euler-Maruyama scheme.
Compute the auxiliary process $Y_t$ (the derivative of the flow) and the Malliavin weights $\Pi$ .
Use backward induction: At each time step $t_k$ , estimate the continuation value $E[\text{Future Payoff} | X_{t_k} = \alpha]$ using the weighted expectation formula.
Determine the option price by comparing the immediate payoff with the estimated continuation value.

3. Key Contributions

Novel Representation Formula: The first derivation of explicit conditional expectation formulas for mean-field SDEs with jumps using Malliavin calculus on Poisson space. Previous works focused on pure diffusion or non-mean-field jump models.
Directional Derivatives in Poisson Space: The systematic application of directional differentiation with respect to the Poisson process in a mean-field setting, which is a new technique in this domain.
Variance-Reduced Estimator: The integration of optimal localization functions (Laplace density) into the mean-field jump framework, providing a robust estimator for high-dimensional problems.
Rigorous Proofs: The paper provides rigorous proofs for the Malliavin differentiability of the solution and the validity of the representation formulas under specific Lipschitz and integrability conditions.

4. Results and Numerical Experiments

The authors validate their method through extensive numerical experiments, comparing the MMC method against Finite Difference (FD) benchmarks.

Test Cases:
- Example 1: Single asset with uniform jumps.
- Example 2: Single asset with Kou double-exponential jumps.
- Example 3: Two-asset portfolio optimization with mixed jump distributions (Uniform and Kou).
Performance Metrics:
- Accuracy: The MMC estimates converge to the Finite Difference benchmarks as the number of simulations ( $M$ ) and time discretization steps ( $N$ ) increase.
- Error Reduction: Plot analysis shows that absolute error decreases significantly with increased discretization and sample size.
- Stability: The use of localization and control variates results in stable estimates with narrow error bars, even with moderate sample sizes ( $M=1000$ ).
Specific Findings:
- The method successfully handles the "early-exercise" boundary of American options.
- In the portfolio optimization example, the method effectively captures the interdependencies between assets driven by the mean-field term $E[X_t]$ .

5. Significance

Theoretical Advancement: This work bridges the gap between Mean-Field Game theory, Jump-Diffusion models, and Malliavin Calculus, providing a unified mathematical tool for complex financial systems.
Practical Application: It offers a viable solution for pricing high-dimensional American options in markets characterized by systemic risk (mean-field) and sudden shocks (jumps), scenarios where traditional grid-based methods are computationally impossible.
Efficiency: By avoiding the "curse of dimensionality" inherent in grid methods and reducing the variance of Monte Carlo estimators, the proposed MMC algorithm enables more accurate and faster pricing of complex derivatives in modern financial engineering.

In conclusion, Sojudi and Tahmasebi present a powerful, mathematically rigorous, and numerically efficient framework for solving conditional expectation problems in mean-field jump-diffusion systems, with immediate applicability to the pricing of American options and portfolio optimization.